Agent CLI Reference¶

Overview¶

The lango agent command group covers two surfaces:

• inspection commands such as status, list, tools, and hooks
• diagnostics commands such as trace, graph, and trace-derived metrics

Inspection-oriented commands are documented in Agent & Memory. This page focuses on the diagnostics surface that uses bootstrap-backed trace storage.

All diagnostics commands accept --output table|json and reject unknown values before bootstrap-dependent work begins. Human-readable and JSON output are routed through the Cobra command writer so wrappers and test harnesses can capture them directly.

Commands¶

lango agent trace list¶

List recent turn traces.

lango agent trace list [--session <key>] [--outcome <outcome>] [--limit N] [--output table|json]

Flags:

lango agent trace show <trace-id>¶

Show a detailed event timeline for one trace.

lango agent trace show <trace-id> [--output table|json]

Flags:

lango agent trace metrics¶

Show trace-derived per-agent performance metrics. This is distinct from lango metrics agents, which reports token usage rather than trace outcomes and durations.

lango agent trace metrics [--agent <name>] [--output table|json]

Flags:

lango agent graph <session-key>¶

Show the delegation graph for a session.

lango agent graph <session-key> [--output table|json]

Flags:

Notes¶

• trace, graph, and trace-derived metrics use bootstrap-backed storage rather than config-only inspection.
• lango agent trace metrics lives under the trace command tree; it is not exposed as a root-level lango agent metrics command.
• If no diagnostics data exists, the commands render explicit empty-state output instead of failing silently.